{
  "api_version": "1",
  "updated": "2026-10-26",
  "latest_event": "2026-10-24",
  "rights": "© Frontier Log, all rights reserved.",
  "model": {
    "slug": "glm-5",
    "name": "GLM-5",
    "lab": "zai",
    "date": "2026-02-12",
    "flagship": false,
    "event": "https://frontierlog.ai/event/2026-02-12-glm-5",
    "releases": [],
    "variant_of": null,
    "made_by": "Z.ai",
    "url": "https://frontierlog.ai/model/glm-5",
    "results": [
      {
        "id": "glm-5:hle:no-tools:1",
        "model": "glm-5",
        "model_name": "GLM-5",
        "model_date": "2026-02-12",
        "event": "https://frontierlog.ai/event/2026-02-12-glm-5",
        "benchmark": "hle",
        "benchmark_name": "Humanity's Last Exam",
        "setting": "no-tools",
        "setting_name": "No tools",
        "score": 30.5,
        "unit": "%",
        "printed": "30.5",
        "reported_by": "zai",
        "run_by": null,
        "label": "GLM-5",
        "source": "https://z.ai/blog/glm-5",
        "source_title": "GLM-5: From Vibe Coding to Agentic Engineering",
        "source_date": "2026-02-12",
        "where": "benchmark table, row 'Humanity's Last Exam', column 'GLM-5 (Thinking)'",
        "own": true,
        "primary": true,
        "parallel": false,
        "notes": "text-only subset; GPT-5.2 judge; up to 131,072 output tokens",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-19",
        "check": "human",
        "corroborated": false,
        "corroborated_on": null,
        "corroborated_by": [],
        "source_capture": "https://frontierlog.ai/captures/8709485a935545d0.png"
      },
      {
        "id": "glm-5:hle:with-tools:1",
        "model": "glm-5",
        "model_name": "GLM-5",
        "model_date": "2026-02-12",
        "event": "https://frontierlog.ai/event/2026-02-12-glm-5",
        "benchmark": "hle",
        "benchmark_name": "Humanity's Last Exam",
        "setting": "with-tools",
        "setting_name": "With tools",
        "score": 50.4,
        "unit": "%",
        "printed": "50.4",
        "reported_by": "zai",
        "run_by": null,
        "label": "GLM-5",
        "source": "https://z.ai/blog/glm-5",
        "source_title": "GLM-5: From Vibe Coding to Agentic Engineering",
        "source_date": "2026-02-12",
        "where": "benchmark table, row 'Humanity's Last Exam w/ Tools', column 'GLM-5 (Thinking)'",
        "own": true,
        "primary": true,
        "parallel": false,
        "notes": "text-only subset; GPT-5.2 judge; 202,752-token context",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-19",
        "check": "human",
        "corroborated": true,
        "corroborated_on": "2026-10-19",
        "corroborated_by": [
          {
            "url": "https://z.ai/blog/glm-5.1",
            "name": "GLM-5.1",
            "title": "GLM-5.1: Towards Long-Horizon Tasks",
            "lab": "zai",
            "date": "2026-04-07",
            "printed": "50.4",
            "where": "Benchmark table, row 'HLE w/ Tools (* = full set for Opus 4.6, Gemini 3.1 Pro, GPT-5.4)', column 'GLM-5'"
          }
        ],
        "source_capture": "https://frontierlog.ai/captures/8709485a935545d0.png"
      },
      {
        "id": "glm-5:gpqa-diamond:default:1",
        "model": "glm-5",
        "model_name": "GLM-5",
        "model_date": "2026-02-12",
        "event": "https://frontierlog.ai/event/2026-02-12-glm-5",
        "benchmark": "gpqa-diamond",
        "benchmark_name": "GPQA Diamond",
        "setting": "default",
        "setting_name": "",
        "score": 86.0,
        "unit": "%",
        "printed": "86.0",
        "reported_by": "zai",
        "run_by": null,
        "label": "GLM-5",
        "source": "https://z.ai/blog/glm-5",
        "source_title": "GLM-5: From Vibe Coding to Agentic Engineering",
        "source_date": "2026-02-12",
        "where": "benchmark table, row 'GPQA-Diamond', column 'GLM-5 (Thinking)'",
        "own": true,
        "primary": true,
        "parallel": false,
        "notes": "up to 131,072 output tokens",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-19",
        "check": "human",
        "corroborated": true,
        "corroborated_on": "2026-10-19",
        "corroborated_by": [
          {
            "url": "https://z.ai/blog/glm-5.1",
            "name": "GLM-5.1",
            "title": "GLM-5.1: Towards Long-Horizon Tasks",
            "lab": "zai",
            "date": "2026-04-07",
            "printed": "86.0",
            "where": "Benchmark table, row 'GPQA-Diamond', column 'GLM-5'"
          }
        ],
        "source_capture": "https://frontierlog.ai/captures/8709485a935545d0.png"
      },
      {
        "id": "glm-5:swe-bench-verified:default:1",
        "model": "glm-5",
        "model_name": "GLM-5",
        "model_date": "2026-02-12",
        "event": "https://frontierlog.ai/event/2026-02-12-glm-5",
        "benchmark": "swe-bench-verified",
        "benchmark_name": "SWE-bench Verified",
        "setting": "default",
        "setting_name": "",
        "score": 77.8,
        "unit": "%",
        "printed": "77.8",
        "reported_by": "zai",
        "run_by": null,
        "label": "GLM-5",
        "source": "https://z.ai/blog/glm-5",
        "source_title": "GLM-5: From Vibe Coding to Agentic Engineering",
        "source_date": "2026-02-12",
        "where": "benchmark table, row 'SWE-bench Verified', column 'GLM-5 (Thinking)'",
        "own": true,
        "primary": true,
        "parallel": false,
        "notes": "OpenHands with a tailored prompt; 200K context",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-19",
        "check": "human",
        "corroborated": true,
        "corroborated_on": "2026-10-19",
        "corroborated_by": [
          {
            "url": "https://qwen.ai/blog?id=qwen3.6",
            "name": "Qwen3.6-Plus",
            "title": "Qwen3.6-Plus: Towards Real World Agents",
            "lab": "alibaba",
            "date": "2026-04-02",
            "printed": "77.8",
            "where": "Score chart image (qwen3.6_plus_score.png), panel SWE-bench Verified, value label on the GLM5 bar (bars identified by legend colour and logo)"
          }
        ],
        "source_capture": "https://frontierlog.ai/captures/8709485a935545d0.png"
      },
      {
        "id": "glm-5:terminal-bench-2-0:default:1",
        "model": "glm-5",
        "model_name": "GLM-5",
        "model_date": "2026-02-12",
        "event": "https://frontierlog.ai/event/2026-02-12-glm-5",
        "benchmark": "terminal-bench-2-0",
        "benchmark_name": "Terminal-Bench 2.0",
        "setting": "default",
        "setting_name": "",
        "score": 56.2,
        "unit": "%",
        "printed": "56.2",
        "reported_by": "zai",
        "run_by": null,
        "label": "GLM-5",
        "source": "https://z.ai/blog/glm-5",
        "source_title": "GLM-5: From Vibe Coding to Agentic Engineering",
        "source_date": "2026-02-12",
        "where": "benchmark table, row 'Terminal-Bench 2.0 Terminus-2', column 'GLM-5 (Thinking)'",
        "own": true,
        "primary": true,
        "parallel": false,
        "notes": "Terminus 2 harness; 2h timeout; 128K context (60.7 on Z.ai's verified version of the benchmark)",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-19",
        "check": "human",
        "corroborated": true,
        "corroborated_on": "2026-10-19",
        "corroborated_by": [
          {
            "url": "https://qwen.ai/blog?id=qwen3.6",
            "name": "Qwen3.6-Plus",
            "title": "Qwen3.6-Plus: Towards Real World Agents",
            "lab": "alibaba",
            "date": "2026-04-02",
            "printed": "56.2",
            "where": "Score chart image (qwen3.6_plus_score.png), panel Terminal-Bench 2.0, value label on the GLM5 bar (bars identified by legend colour and logo)"
          },
          {
            "url": "https://z.ai/blog/glm-5.1",
            "name": "GLM-5.1",
            "title": "GLM-5.1: Towards Long-Horizon Tasks",
            "lab": "zai",
            "date": "2026-04-07",
            "printed": "56.2",
            "where": "Benchmark table, row 'Terminal-Bench 2.0 Terminus-2', column 'GLM-5'"
          }
        ],
        "source_capture": "https://frontierlog.ai/captures/8709485a935545d0.png"
      },
      {
        "id": "glm-5:terminal-bench-2-0:default:2",
        "model": "glm-5",
        "model_name": "GLM-5",
        "model_date": "2026-02-12",
        "event": "https://frontierlog.ai/event/2026-02-12-glm-5",
        "benchmark": "terminal-bench-2-0",
        "benchmark_name": "Terminal-Bench 2.0",
        "setting": "default",
        "setting_name": "",
        "score": 56.2,
        "unit": "%",
        "printed": "56.2",
        "reported_by": "zai",
        "run_by": null,
        "label": "GLM-5",
        "source": "https://z.ai/blog/glm-5",
        "source_title": "GLM-5: From Vibe Coding to Agentic Engineering",
        "source_date": "2026-02-12",
        "where": "benchmark table, row 'Terminal-Bench 2.0 Claude Code', column 'GLM-5 (Thinking)'",
        "own": true,
        "primary": false,
        "parallel": false,
        "notes": "Claude Code harness, thinking; no time limit; averaged over 5 runs (61.1 on Z.ai's verified version of the benchmark)",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-19",
        "check": "human",
        "corroborated": true,
        "corroborated_on": "2026-10-19",
        "corroborated_by": [
          {
            "url": "https://qwen.ai/blog?id=qwen3.6",
            "name": "Qwen3.6-Plus",
            "title": "Qwen3.6-Plus: Towards Real World Agents",
            "lab": "alibaba",
            "date": "2026-04-02",
            "printed": "56.2",
            "where": "Score chart image (qwen3.6_plus_score.png), panel Terminal-Bench 2.0, value label on the GLM5 bar (bars identified by legend colour and logo)"
          },
          {
            "url": "https://z.ai/blog/glm-5.1",
            "name": "GLM-5.1",
            "title": "GLM-5.1: Towards Long-Horizon Tasks",
            "lab": "zai",
            "date": "2026-04-07",
            "printed": "56.2",
            "where": "Benchmark table, row 'Terminal-Bench 2.0 Terminus-2', column 'GLM-5'"
          }
        ],
        "source_capture": "https://frontierlog.ai/captures/8709485a935545d0.png"
      },
      {
        "id": "glm-5:browsecomp:default:1",
        "model": "glm-5",
        "model_name": "GLM-5",
        "model_date": "2026-02-12",
        "event": "https://frontierlog.ai/event/2026-02-12-glm-5",
        "benchmark": "browsecomp",
        "benchmark_name": "BrowseComp",
        "setting": "default",
        "setting_name": "",
        "score": 62.0,
        "unit": "%",
        "printed": "62.0",
        "reported_by": "zai",
        "run_by": null,
        "label": "GLM-5",
        "source": "https://z.ai/blog/glm-5",
        "source_title": "GLM-5: From Vibe Coding to Agentic Engineering",
        "source_date": "2026-02-12",
        "where": "benchmark table, row 'BrowseComp', column 'GLM-5 (Thinking)'",
        "own": true,
        "primary": true,
        "parallel": false,
        "notes": "without context management: keeps the last 5 turns",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-19",
        "check": "human",
        "corroborated": true,
        "corroborated_on": "2026-10-19",
        "corroborated_by": [
          {
            "url": "https://z.ai/blog/glm-5.1",
            "name": "GLM-5.1",
            "title": "GLM-5.1: Towards Long-Horizon Tasks",
            "lab": "zai",
            "date": "2026-04-07",
            "printed": "62.0",
            "where": "Benchmark table, row 'BrowseComp (no context management)', column 'GLM-5'"
          }
        ],
        "source_capture": "https://frontierlog.ai/captures/8709485a935545d0.png"
      },
      {
        "id": "glm-5:browsecomp:default:2",
        "model": "glm-5",
        "model_name": "GLM-5",
        "model_date": "2026-02-12",
        "event": "https://frontierlog.ai/event/2026-02-12-glm-5",
        "benchmark": "browsecomp",
        "benchmark_name": "BrowseComp",
        "setting": "default",
        "setting_name": "",
        "score": 75.9,
        "unit": "%",
        "printed": "75.9",
        "reported_by": "zai",
        "run_by": null,
        "label": "GLM-5",
        "source": "https://z.ai/blog/glm-5",
        "source_title": "GLM-5: From Vibe Coding to Agentic Engineering",
        "source_date": "2026-02-12",
        "where": "benchmark table, row 'BrowseComp w/ Context Manage', column 'GLM-5 (Thinking)'",
        "own": true,
        "primary": false,
        "parallel": false,
        "notes": "with context management, discard-all strategy",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-19",
        "check": "human",
        "corroborated": true,
        "corroborated_on": "2026-10-19",
        "corroborated_by": [
          {
            "url": "https://z.ai/blog/glm-5.1",
            "name": "GLM-5.1",
            "title": "GLM-5.1: Towards Long-Horizon Tasks",
            "lab": "zai",
            "date": "2026-04-07",
            "printed": "75.9",
            "where": "Benchmark table, row 'BrowseComp w/ Context Manage', column 'GLM-5'"
          }
        ],
        "source_capture": "https://frontierlog.ai/captures/8709485a935545d0.png"
      },
      {
        "id": "glm-5:mcp-atlas:default:1",
        "model": "glm-5",
        "model_name": "GLM-5",
        "model_date": "2026-02-12",
        "event": "https://frontierlog.ai/event/2026-02-12-glm-5",
        "benchmark": "mcp-atlas",
        "benchmark_name": "MCP Atlas",
        "setting": "default",
        "setting_name": "",
        "score": 67.8,
        "unit": "%",
        "printed": "67.8",
        "reported_by": "zai",
        "run_by": null,
        "label": "GLM-5",
        "source": "https://z.ai/blog/glm-5",
        "source_title": "GLM-5: From Vibe Coding to Agentic Engineering",
        "source_date": "2026-02-12",
        "where": "benchmark table, row 'MCP-Atlas Public Set', column 'GLM-5 (Thinking)'",
        "own": true,
        "primary": true,
        "parallel": false,
        "notes": "500-task public set; thinking mode; 10-minute timeout per task; Gemini 3 Pro judge",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-19",
        "check": "human",
        "corroborated": false,
        "corroborated_on": null,
        "corroborated_by": [],
        "source_capture": "https://frontierlog.ai/captures/8709485a935545d0.png"
      },
      {
        "id": "glm-5:swe-bench-pro:default:1",
        "model": "glm-5",
        "model_name": "GLM-5",
        "model_date": "2026-02-12",
        "event": "https://frontierlog.ai/event/2026-02-12-glm-5",
        "benchmark": "swe-bench-pro",
        "benchmark_name": "SWE-bench Pro",
        "setting": "default",
        "setting_name": "",
        "score": 55.1,
        "unit": "%",
        "printed": "55.1",
        "reported_by": "zai",
        "run_by": null,
        "label": "GLM-5",
        "source": "https://z.ai/blog/glm-5.1",
        "source_title": "GLM-5.1: Towards Long-Horizon Tasks",
        "source_date": "2026-04-07",
        "where": "benchmark table near the end, row 'SWE-Bench Pro', column 'GLM-5'",
        "own": false,
        "primary": true,
        "parallel": false,
        "notes": "OpenHands with a tailored prompt",
        "caveat": "The source doesn't say whether this is the public or the commercial set",
        "human_verified": true,
        "human_verified_on": "2026-10-19",
        "check": "human",
        "corroborated": true,
        "corroborated_on": "2026-10-19",
        "corroborated_by": [
          {
            "url": "https://qwen.ai/blog?id=qwen3.6",
            "name": "Qwen3.6-Plus",
            "title": "Qwen3.6-Plus: Towards Real World Agents",
            "lab": "alibaba",
            "date": "2026-04-02",
            "printed": "55.1",
            "where": "Score chart image (qwen3.6_plus_score.png), panel SWE-bench Pro, value label on the GLM5 bar (bars identified by legend colour and logo)"
          }
        ],
        "source_capture": "https://frontierlog.ai/captures/35c66845b36400f8.png"
      },
      {
        "id": "glm-5:mcp-atlas:default:2",
        "model": "glm-5",
        "model_name": "GLM-5",
        "model_date": "2026-02-12",
        "event": "https://frontierlog.ai/event/2026-02-12-glm-5",
        "benchmark": "mcp-atlas",
        "benchmark_name": "MCP Atlas",
        "setting": "default",
        "setting_name": "",
        "score": 69.2,
        "unit": "%",
        "printed": "69.2",
        "reported_by": "zai",
        "run_by": null,
        "label": "GLM-5",
        "source": "https://z.ai/blog/glm-5.1",
        "source_title": "GLM-5.1: Towards Long-Horizon Tasks",
        "source_date": "2026-04-07",
        "where": "benchmark table near the end, row 'MCP-Atlas Public Set', column 'GLM-5'",
        "own": false,
        "primary": false,
        "parallel": false,
        "notes": "500-task public set; Gemini 3 Pro judge",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-19",
        "check": "human",
        "corroborated": false,
        "corroborated_on": null,
        "corroborated_by": [],
        "source_capture": "https://frontierlog.ai/captures/35c66845b36400f8.png"
      }
    ]
  }
}